Publicação
Double Conversion Uninterruptible Power Supply with Internal Fault Diagnosis and Tolerance, based on a Multilevel Topology
| Resumo: | The main objective of this work is the study and development of a highly reliable double conversion uninterruptible power supply (UPS) with fully-automated fault diagnosis and fault-tolerant measures. A multilevel topology using two 3-level Neutral-Point-Clamped (3LNPC) converters and a 3-level DC-DC (3LDC) converter was adopted for its benefits in converter performance and fault-tolerant capabilities. Due to its advantages, finite control set model predictive control (FCS-MPC) is adopted for the control of all converters in the UPS. A cooperative MPC technique is presented to allow greater collaboration between the 3 converters and improve the overall performance of the UPS without increasing the computational load of the controller. A new dynamically weighted MPC (DWMPC) control strategy is also proposed for all converter in the UPS. This technique provides significantly increased system stability, while improving steady-state performance and allowing a more effective balance between multiple objective functions, compared to the conventional FCS-MPC formulation. Despite a higher complexity, this algorithm can provide more intuitive weighting factor selection and does not significantly increase the required amount of calculations, especially in complex systems, such as multilevel converters. A complete prototype of the proposed UPS was designed and built during this work, including all necessary power converters, small signal electronics, etc. The performance of the developed UPS is evaluated in this document using simulation and experimental results, demonstrating the advantages of the proposed techniques. The results shown that a very stable UPS is obtained, with great steady-state performance and a fast dynamic response. The impact of semiconductor OC faults in the operation of the UPS is also studied in detail, both theoretically and experimentally. This study confirms the critical impact that OC faults can have on the system and reiterates the importance of including fault diagnosis and fault correction measures in UPS systems. The proposed DWMPC is also shown to be advantageous in the presence of faults. The analysis of OC faults provided the basis for the development of several real-time fault diagnosis algorithms, presented in this work. A single-fault diagnosis algorithm is first proposed for NPC converters, followed by a more complex multiple fault diagnostic solution. Then, a multiple fault diagnosis algorithm based on the same diagnostic principle is proposed for the 3LDC converter. These algorithms present several advantages compared to those available in the literature. Even tough almost all solutions in the literature focus exclusively on OC IGBT faults, diode faults can also have a severe impact on the power converter. For this reason, all developed algorithms presented in this thesis consider OC faults in all semiconductors in the topology, including all IGBTs and diodes. Furthermore, most diagnostic solutions found in the literature consider only the possibility of a single OC fault, or a maximum of two simultaneous faults. The proposed multiple fault diagnosis algorithms (for 3LNPC and 3LDC) can detect any number of faults in all IGBTs and diodes in the topology, including intermittent faults. This makes them extremely versatile and reliable. The proposed diagnostic algorithms are independent from the load, can be used in a large variety of applications and require no additional sensors. These algorithms were tested experimentally using a microprocessor and an FPGA, displaying great robustness and extremely fast diagnostics (in tens or hundreds of microseconds, or in as little as a few microseconds, when executed in the FPGA). Fault diagnosis was tested with up to 4 simultaneous faults in the same converter, delivering fast and accurate fault identifications, even in unfavorable conditions. Two fault-tolerant strategies are then presented and analyzed: one with a non-redundant topology and one with a redundant converter leg. These strategies are based on techniques commonly used in 3-phase converters, but significant improvements are proposed, regarding both the reconfiguration procedure and the controller behavior. Furthermore, these fault-tolerant principles are transposed to the 3LDC converter case, in order to allow fault-tolerance in this converter in similar conditions (this was previously unstudied). Both fault-tolerant solutions provide full-rated post-fault operation - critical in the UPS. The advantages and drawbacks of each technique are discussed in detail, considering the requirements of the UPS and the protection of the critical load. The operation of these methods is demonstrated and analyzed experimentally, as well as the performance reduction after the fault. The scalability of fault tolerance to high-power modular UPS systems is also briefly discussed. |
|---|---|
| Autores principais: | Caseiro, Luís Miguel Antunes |
| Assunto: | Uninterruptible Power Supply Multilevel Converters Fault Diagnosis Fault Tolerance Predictive Control |
| Ano: | 2017 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Coimbra |
| Idioma: | inglês |
| Origem: | Estudo Geral - Universidade de Coimbra |
| Resumo: | The main objective of this work is the study and development of a highly reliable double conversion uninterruptible power supply (UPS) with fully-automated fault diagnosis and fault-tolerant measures. A multilevel topology using two 3-level Neutral-Point-Clamped (3LNPC) converters and a 3-level DC-DC (3LDC) converter was adopted for its benefits in converter performance and fault-tolerant capabilities. Due to its advantages, finite control set model predictive control (FCS-MPC) is adopted for the control of all converters in the UPS. A cooperative MPC technique is presented to allow greater collaboration between the 3 converters and improve the overall performance of the UPS without increasing the computational load of the controller. A new dynamically weighted MPC (DWMPC) control strategy is also proposed for all converter in the UPS. This technique provides significantly increased system stability, while improving steady-state performance and allowing a more effective balance between multiple objective functions, compared to the conventional FCS-MPC formulation. Despite a higher complexity, this algorithm can provide more intuitive weighting factor selection and does not significantly increase the required amount of calculations, especially in complex systems, such as multilevel converters. A complete prototype of the proposed UPS was designed and built during this work, including all necessary power converters, small signal electronics, etc. The performance of the developed UPS is evaluated in this document using simulation and experimental results, demonstrating the advantages of the proposed techniques. The results shown that a very stable UPS is obtained, with great steady-state performance and a fast dynamic response. The impact of semiconductor OC faults in the operation of the UPS is also studied in detail, both theoretically and experimentally. This study confirms the critical impact that OC faults can have on the system and reiterates the importance of including fault diagnosis and fault correction measures in UPS systems. The proposed DWMPC is also shown to be advantageous in the presence of faults. The analysis of OC faults provided the basis for the development of several real-time fault diagnosis algorithms, presented in this work. A single-fault diagnosis algorithm is first proposed for NPC converters, followed by a more complex multiple fault diagnostic solution. Then, a multiple fault diagnosis algorithm based on the same diagnostic principle is proposed for the 3LDC converter. These algorithms present several advantages compared to those available in the literature. Even tough almost all solutions in the literature focus exclusively on OC IGBT faults, diode faults can also have a severe impact on the power converter. For this reason, all developed algorithms presented in this thesis consider OC faults in all semiconductors in the topology, including all IGBTs and diodes. Furthermore, most diagnostic solutions found in the literature consider only the possibility of a single OC fault, or a maximum of two simultaneous faults. The proposed multiple fault diagnosis algorithms (for 3LNPC and 3LDC) can detect any number of faults in all IGBTs and diodes in the topology, including intermittent faults. This makes them extremely versatile and reliable. The proposed diagnostic algorithms are independent from the load, can be used in a large variety of applications and require no additional sensors. These algorithms were tested experimentally using a microprocessor and an FPGA, displaying great robustness and extremely fast diagnostics (in tens or hundreds of microseconds, or in as little as a few microseconds, when executed in the FPGA). Fault diagnosis was tested with up to 4 simultaneous faults in the same converter, delivering fast and accurate fault identifications, even in unfavorable conditions. Two fault-tolerant strategies are then presented and analyzed: one with a non-redundant topology and one with a redundant converter leg. These strategies are based on techniques commonly used in 3-phase converters, but significant improvements are proposed, regarding both the reconfiguration procedure and the controller behavior. Furthermore, these fault-tolerant principles are transposed to the 3LDC converter case, in order to allow fault-tolerance in this converter in similar conditions (this was previously unstudied). Both fault-tolerant solutions provide full-rated post-fault operation - critical in the UPS. The advantages and drawbacks of each technique are discussed in detail, considering the requirements of the UPS and the protection of the critical load. The operation of these methods is demonstrated and analyzed experimentally, as well as the performance reduction after the fault. The scalability of fault tolerance to high-power modular UPS systems is also briefly discussed. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.